Color-Spatial AutoAugment: Automated Image Augmentation and Policy Optimization

Data augmentation is a highly effective technique for improving modern image classifiers’ accuracy and has continuously improved over the years. In this paper, We improve upon the previously published approach to implementing AutoAugment policies known as "Color-Spatial AutoAugment"; our implementation utilizes the best policies discovered for specific datasets. It categorizes them into color and spatial, thereby improving image classification accuracy. Applied to the CIFAR-10 dataset, our method significantly improved the performance of the networks in self-supervised training (91.13% accuracy and 0.2837 loss). We share those findings to highlight the potential of Color-Spatial AutoAugment and the improvement over similar augmentation methods.